Automated Anomaly Detection and Predictive Maintenance in Linear Reciprocating Sliding Wear Tests via a Hybrid Bayesian-Gaussian Process Framework

**Abstract:** This paper proposes a novel framework for automated anomaly detection and predictive maintenance in linear reciprocating sliding wear tests, a critical process in materials science and tribology. Current analysis relies heavily on manual inspection and statistical averaging, limiting accuracy, speed, and predictive capabilities. Our approach, combining a Bayesian network for feature engineering with a Gaussian process regression (GPR) for predictive modeling, enables real-time anomaly identification and trajectory forecasting of wear parameters, facilitating proactive maintenance and optimized testing protocols. The framework is demonstrably viable for immediate commercial deployment, offering enhanced efficiency and reliability in materials research and development.